Step-by-Step Guide to GPU Setup for Machine Learning & Deep Learning!

Hello everyone, and welcome to this tutorial on how to use your GPU for AI tasks, including machine learning and deep learning, instead of relying on your CPU. Using a GPU can significantly accelerate computations, especially for deep learning models. Let's get started step by step.

Step 1: Create a New Environment and Install Required Packages

First, open the Anaconda Prompt and create a new environment with TensorFlow GPU support. Use the following command:

conda create –n tf-GPU tensorflow-gpu

This will create a new environment named tf-GPU with the necessary TensorFlow GPU version installed.

Step 2: Activate the New Environment

Activate the newly created environment by typing:

conda activate tf-GPU

This ensures all your installations and tasks are isolated to this environment.

Step 3: Install TensorFlow 2.10

TensorFlow 2.10 is compatible with many GPUs. Install it using:

python -m pip install "tensorflow==2.10.0"

Step 4: Install the Jupyter Kernel for This Environment

To use this environment in Jupyter Notebook, run the following commands:

pip install ipykernel

python -m ipykernel install --user --name=tf-GPU --display-name "Python (tf-GPU)"

This makes the tf-GPU environment available as a kernel in Jupyter Notebook.

Step 5: Open Jupyter Notebook

Now, open Jupyter Notebook:

jupyter notebook

In Jupyter, go to the Kernel menu, select Change Kernel, and choose Python (tf-GPU).

Step 6: Verify GPU Availability in Jupyter Notebook

To check if TensorFlow recognizes your GPU, run the following code:

import tensorflow as tf

print("GPU Details:", tf.config.list_physical_devices('GPU'))

This should display details of your GPU.

Step 7: First Example - Comparing CPU and GPU Performance

Here’s a simple example to compare the time taken for matrix multiplication on GPU and CPU:

import tensorflow as tf
import time

# Check available GPU
print("Available GPU:", tf.config.list_physical_devices('GPU'))

# Create large random matrices
matrix_size = 10000
a = tf.random.normal((matrix_size, matrix_size))
b = tf.random.normal((matrix_size, matrix_size))

# Perform matrix multiplication on GPU
start_time = time.time()
c = tf.matmul(a, b)
end_time = time.time()
print(f"Matrix multiplication on GPU completed in {end_time - start_time:.4f} seconds")

# Perform matrix multiplication on CPU
with tf.device('/CPU:0'):
    start_time = time.time()
    c_cpu = tf.matmul(a, b)
    end_time = time.time()
    print(f"Matrix multiplication on CPU completed in {end_time - start_time:.4f} seconds")

Step 8: Second Example - Training a Neural Network Using the GPU

Let’s now train a simple neural network on the MNIST dataset using the GPU:

import tensorflow as tf
from tensorflow.keras.datasets import mnist
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Flatten
from tensorflow.keras.optimizers import Adam

# Load MNIST dataset
(x_train, y_train), (x_test, y_test) = mnist.load_data()

# Normalize the data (scale pixel values to 0-1)
x_train, x_test = x_train / 255.0, x_test / 255.0

# Build a simple neural network model
model = Sequential([
    Flatten(input_shape=(28, 28)),  # Flatten 28x28 images into 1D vectors
    Dense(128, activation='relu'), # First dense layer with 128 units
    Dense(10, activation='softmax') # Output layer with 10 classes (digits 0-9)
])

# Compile the model
model.compile(optimizer=Adam(),
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])

# Train the model
model.fit(x_train, y_train, epochs=10, batch_size=32, validation_data=(x_test, y_test))

# Evaluate the model on the test set
test_loss, test_accuracy = model.evaluate(x_test, y_test, verbose=2)
print(f"Test Accuracy: {test_accuracy:.4f}")

Summary

In this tutorial, we learned how to set up a GPU-enabled TensorFlow environment, verify GPU usage, and perform computations using TensorFlow. We also compared GPU and CPU performance and trained a simple neural network using the GPU. By leveraging GPUs, you can significantly speed up your AI tasks and focus on building better models faster.

For a more comprehensive learning experience, check out my Udemy course: AI for Geospatial Analysis. This course dives deeper into AI and machine learning applications for geospatial data analysis.

Step-by-Step Guide to GPU Setup for Machine Learning & Deep Learning!

Did you find this article valuable?